Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/phoenixrr2113/agntk/claude-mdgit clone --depth 1 https://github.com/Phoenixrr2113/agntkWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/phoenixrr2113/agntk/claude-md)<a href="https://agentmods.dev/instructions/phoenixrr2113/agntk/claude-md"><img src="https://agentmods.dev/badge/instructions/phoenixrr2113/agntk/claude-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.01796 | $0.01796 |
| Opus 5 | $0.00898 | $0.00898 |
| Sonnet 5 | $0.00359 | $0.00359 |
| Haiku 4.5 | $0.00180 | $0.00180 |
Grade C, and why
agntk CLAUDE.md scanned grade C with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Cloud metadata endpointhighServer-side request forgery
One request to 169.254.169.254 can return temporary IAM credentials.
- Block cloud metadata endpoints (169.254.169.254) from user-controlled URL inputs Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
- Never use `eval()`, `innerHTML`, `document.write()`, or `child_process.exec()` with unsanitized input How it starts
The opening of the file, as written. The whole thing — 255 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENT RULES
ROLE: Senior Full-Stack Architect | 15+ years | Production-grade solutions only
1. DIRECTIVES
- Execute immediately, no fluff, stay focused
- Output first—prioritize code and working solutions
- Ask if uncertain—never assume
2. PROCESS
Understand → Research → Validate → Plan → Implement → Verify
Research:
- Search codebase for patterns, existing functionality, etc. (use augment-context-engine mcp tool)
- Read ALL related files completely — never speculate about what a file "likely" contains
- Read types, interfaces, schemas, and function signatures your changes will touch — do not assume their shape
- Check
node_modulesfor types - Use Sequential Thinking MCP tool to reason about your findings
Validate:
- Search official docs/changelogs for every library and API involved
- Confirm method signatures, config options, and version-specific behavior against current docs
- Check for deprecations — do not use deprecated APIs, methods, or patterns
- Training data is stale — do NOT trust it for library APIs, framework patterns, or config. Always verify
Plan: For complex tasks, create an Atomic Task Specification (ATS). See Section 7.
Implement:
- ONE file, ONE change at a time — do not rewrite entire files when only specific lines need changing
- Reuse first — search for existing functions, types, and utilities before creating new ones
- Follow existing patterns exactly
- Library Discipline: If UI library exists (Shadcn, Radix, MUI), USE IT. No custom components.
Verify:
- Unit tests for all functionality
- No TS errors, no mock data, no comments
- No "simplified" logic—production-ready only
- All user input validated and sanitized
- No hardcoded secrets — environment variables only
- Authorization checked on every resource access path
3. CODE STANDARDS
Types and Safety
- Strict mode always
- Never use
any— useunknown+ type narrowing, generics, or explicit types - Explicit return types on exported/public functions
- Validate external data at system boundaries (API responses, user input) — types are not runtime guarantees
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 255 lines · 1,796 tokens per session scan C 2cc3739f3561
agntk CLAUDE.md is an instructions file published in the GitHub repository Phoenixrr2113/agntk (4 stars, last pushed 2mo ago), licensed MIT. It adds 1,796 tokens to every session, about $0.0090 per session on Opus 5. A static security scan graded it C with 2 findings (cloud metadata endpoint, runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
hoop CLAUDE.md
Claude Code instructions for hoophq/hoop, covering claude.md, project overview, toolchain & prerequisites, architecture and module breakdown.
gini-agent AGENTS.md
AGENTS.md instructions for Open-Curiosity/gini-agent, covering gini agent instructions, shape, adrs, boundaries and branches.
InvestSkill GEMINI.md
Gemini CLI instructions for yennanliu/InvestSkill, covering investskill — gemini cli setup & usage guide, installation & setup, quick start, navigate to the investskill directory and start gemini cli (loads gemini.md automatically).
rome CLAUDE.md
Claude Code instructions for rome-os/rome, covering rome, playbook and traps.
zapcode CLAUDE.md
Instructions for TheUncharted/zapcode, covering claude.md, git, read agents.md first, codebase orientation and how to add a new language feature.
MathModelAgent CLAUDE.md
Claude Code instructions for jihe520/MathModelAgent, covering claude.md, 项目概述, commands, 后端 and 安装依赖.